Benchmark Almanac

M5HisDoc

TongGuOCR 93.8% tops its leaderboard Qwen3.7-Plus 64.5% leads the models we track
Not in model cards

Models locate text, recognize characters, and recover reading order on Chinese historical pages. The hard subset distorts the same 4,000 pages as the regular subset.

TongGuOCR compares 11 general vision language models on the official 1,000 regular test pages. The authors require institutional approval for dataset access.

Released
June 2023
Kind
Documents, charts and text in images
Level
University
Scored
Checked automatically
Answers
Open-ended
Tasks
Kept private
Measures
Capability
Language
Chinese
In and out
Documents → Text
Interaction
Single turn
Refresh
One fixed set
Run by
Anyone
Strongest model tested
GPT-5.5 #21 on the capability index
Last model added
June 2026
Scores refreshed
November 20th, 2026
34
Entries on TongGuOCR’s leaderboard, 11 models we track

TongGuOCR tops the leaderboard with 93.8% Qwen3.7-Plus leads with 64.5% among models released by June 2026

Best score per entry from TongGuOCR’s leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. The best model we track is Qwen3.7-Plus, at 64.5%. This view shows only models we track, each at its best.

The newest model with a score here came out in June 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 TongGuOCR Page recognition 93.8%
  2. 2 Ma et al. Text-line recognition 92.1%
  3. 3 CRNN Text-line recognition 91.4%
  4. 4 RobustScanner Text-line recognition 90.8%
  5. 5 Liu et al. Page recognition 89.5%
  6. 6 ZCTRN Text-line recognition 88.5%
  7. 7 Peng et al. Text-line recognition 88.3%
  8. 8 ASTER Text-line recognition 87.7%
  9. 9 NRTR Text-line recognition 83%
  10. 10 Qwen3.7-Plus Page recognition 64.5%
  11. 11 Gemini 3.1 Pro Page recognition 61%
  12. 12 HanDoc-OrderOCR Page recognition 54.5%
  13. 13 PaddleOCR-VL-1.6 Page recognition 53.6%
  14. 14 Qwen3.5-9B Page recognition 53%
  15. 15 HunyuanOCR Page recognition 48.9%
  16. 16 GLM-OCR Page recognition 46.7%
  17. 17 Kimi K2.6 Page recognition 45.3%
  18. 18 GPT-5.5 Page recognition 45%
  19. 19 PaddleOCR-VL-1.5 Page recognition 40.3%
  20. 20 MinerU2.5-Pro Page recognition 38.5%
  21. 21 MinerU2.5 Page recognition 24.8%
  22. 22 MiniMax-M3 Page recognition 5.97%
  23. 23 Qianfan-OCR Page recognition 4.58%
  24. 24 Qwen3-VL-8B Page recognition -5.07%
  25. 25 InternVL3.5-8B Page recognition -27.80%
  26. 26 dots.ocr Page recognition -35.40%
  27. 27 Qwen3.5-0.8B Page recognition -63.41%
  28. 28 Qwen3-VL-2B Page recognition -100.41%
  29. 29 InternVL3.5-2B Page recognition -130.28%
  30. 30 MonkeyOCR-3B Page recognition -169.16%
  31. 31 DeepSeek-OCR Page recognition -208.84%
  32. 32 FireRed-OCR Page recognition -244.73%
  33. 33 DeepSeek-OCR 2 Page recognition -357.64%
  34. 34 MonkeyOCR-1.2B Page recognition -422.25%

Source: TongGuOCR’s leaderboard

#EntryScoreSettingDateSource
1 TongGuOCR 93.8% Page recognition Official leaderboard
2 Ma et al. 92.1% Text-line recognition Official leaderboard
3 CRNN 91.4% Text-line recognition Official leaderboard
4 RobustScanner 90.8% Text-line recognition Official leaderboard
5 Liu et al. 89.5% Page recognition Official leaderboard
6 ZCTRN 88.5% Text-line recognition Official leaderboard
7 Peng et al. 88.3% Text-line recognition Official leaderboard
8 ASTER 87.7% Text-line recognition Official leaderboard
9 NRTR 83% Text-line recognition Official leaderboard
10 Qwen3.7-Plus 64.5% Page recognition Jun 2026 Official leaderboard
11 Gemini 3.1 Pro 61% Page recognition Feb 2026 Official leaderboard
12 HanDoc-OrderOCR 54.5% Page recognition Official leaderboard
13 PaddleOCR-VL-1.6 53.6% Page recognition Official leaderboard
14 Qwen3.5-9B 53% Page recognition Feb 2026 Official leaderboard
15 HunyuanOCR 48.9% Page recognition Official leaderboard
16 GLM-OCR 46.7% Page recognition Jan 2026 Official leaderboard
17 Kimi K2.6 45.3% Page recognition Apr 2026 Official leaderboard
18 GPT-5.5 45% Page recognition Apr 2026 Official leaderboard
19 PaddleOCR-VL-1.5 40.3% Page recognition Official leaderboard
20 MinerU2.5-Pro 38.5% Page recognition Official leaderboard
21 MinerU2.5 24.8% Page recognition Official leaderboard
22 MiniMax-M3 5.97% Page recognition Jun 2026 Official leaderboard
23 Qianfan-OCR 4.58% Page recognition Official leaderboard
24 Qwen3-VL-8B -5.07% Page recognition Official leaderboard
25 InternVL3.5-8B -27.80% Page recognition Aug 2025 Official leaderboard
26 dots.ocr -35.40% Page recognition Official leaderboard
27 Qwen3.5-0.8B -63.41% Page recognition Feb 2026 Official leaderboard
28 Qwen3-VL-2B -100.41% Page recognition Official leaderboard
29 InternVL3.5-2B -130.28% Page recognition Official leaderboard
30 MonkeyOCR-3B -169.16% Page recognition Official leaderboard
31 DeepSeek-OCR -208.84% Page recognition Oct 2025 Official leaderboard
32 FireRed-OCR -244.73% Page recognition Official leaderboard
33 DeepSeek-OCR 2 -357.64% Page recognition Jan 2026 Official leaderboard
34 MonkeyOCR-1.2B -422.25% Page recognition Official leaderboard
#ModelScoreSettingDateSource
1 Qwen3.7-Plus 64.5% Page recognition Jun 2026 Official leaderboard
2 Gemini 3.1 Pro 61% Page recognition Feb 2026 Official leaderboard
3 Qwen3.5-9B 53% Page recognition Feb 2026 Official leaderboard
4 GLM-OCR 46.7% Page recognition Jan 2026 Official leaderboard
5 Kimi K2.6 45.3% Page recognition Apr 2026 Official leaderboard
6 GPT-5.5 45% Page recognition Apr 2026 Official leaderboard
7 MiniMax-M3 5.97% Page recognition Jun 2026 Official leaderboard
8 InternVL3.5-8B -27.80% Page recognition Aug 2025 Official leaderboard
9 Qwen3.5-0.8B -63.41% Page recognition Feb 2026 Official leaderboard
10 DeepSeek-OCR -208.84% Page recognition Oct 2025 Official leaderboard
11 DeepSeek-OCR 2 -357.64% Page recognition Jan 2026 Official leaderboard

4 models have held number one on TongGuOCR’s leaderboard since its release in June 2023

Lab claims count from the day the card came out, independent runs from the model’s release.

Shanghai AI Lab Z.ai Google DeepMind Qwen
TongGuOCR leaderboard Not tested since Jun 2026
Qwen3.7-Plus 64.5%

Source: TongGuOCR’s leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On TongGuOCR’s leaderboard Best so far
0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Qwen3.7-Plus Gemini 3.1 Pro GLM-OCR InternVL3.5-8B 0% 20% 40% 60% 80% 100% 2026 Near the ceiling (85%) Qwen3.7-Plus GLM-OCR

Source: TongGuOCR’s leaderboard

No lab model card reports it yet

We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.

More in Vision

All 636 in Vision
  • MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy. Qwen 3.6 Plus 86%
  • MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
  • CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
  • MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy. Seed2.1 Pro 90.7%
  • MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy. Qwen 3.8 Max 97.7%
  • MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy. Qwen 3.8 Max 85.9%

Data audit

All 10 checks pass. Every benchmark

Catalog record 2 pass
  • Pass Task access The tasks are private.
  • Pass Confidence Checked against a primary source.
Scores 4 pass 1 note
  • Pass Independent runs Scores from the benchmark’s paper (34 entries).
  • Pass Models scored 11 models we track, plus 23 other entries on its leaderboard.
  • Pass Newest score The newest is from June 2026, for Qwen3.7-Plus.
  • Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.5 (#21).
  • Pass Headroom The best is 64.5%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on TongGuOCR: 11 rows for models we track and 23 other entries at the last read.
  • Pass Last read November 20th, 2026.
  • Pass Read worked The last read worked.
Consistency 1 pass 1 note
  • Note Lab claims No model has both a lab number and an independent run.
  • Pass Known issues No published audit found problems.

How we check